Category Archives: Thesis

CAPTAIN: Comprehensive Composition Assistance for Photo Taking

Abstract: Many people are interested in taking astonishing photos and sharing with others. Emerging high-tech hardware and software facilitate ubiquitousness and functionality of digital photography. Because composition matters in photography, researchers have leveraged some common composition techniques, such as the rule of thirds, the triangle technique, and the perspective-related techniques, to assess the aesthetic quality of photos computationally. However, composition techniques developed by professionals are far more diverse than well-documented techniques can cover. We leverage the vast under-explored innovations in photography for computational composition assistance, and there is a lack of a holistic framework to capture important aspects of a given scene and help individuals by constructive clues to take a better shot in their adventure. We propose a comprehensive framework, named CAPTAIN (Composition Assistance for Photo Taking), containing integrated deep-learned semantic detectors, sub-genre categorization, artistic pose clustering, personalized aesthetics-based image retrieval, and style set matching. The framework is backed by a large dataset crawled from a photo-sharing Website with mostly photography enthusiasts and professionals.
The work proposes a sequence of steps that have not been explored in the past by researchers.
The work addresses personal preferences for composition through presenting a ranked-list of photographs to the user based on user-specified weights in the similarity measure. We believe our design leveraging user-defined preferences. Our framework extracts ingredients of a given snapshot of a scene (i.e. the scene that the user is interested in taking a picture of) as a set of composition-related features ranging from low-level features such as color, pattern, and texture to high-level features such as pose, category, rating, gender, and object. Our composition model, indexed offline, is used to provide visual exemplars as recommendations for the scene, which is a novel model for aesthetics-related image retrieval. We believe our design leveraging user-defined preferences The matching algorithm recognizes the best shot among a sequence of shots with respect to the user’s preferred style set. We have conducted a number of experiments on the newly proposed components and reported findings. A user study demonstrates that the work is useful to those taking photos.

Keywords: Computational Composition, Image Aesthetics, Photography, Deep Learning, Image Retrieval

captain_springer_ijcv

Deep-learned Models and Photography Idea Retrieval

Intelligent Portrait Composition Assistance (IPCA)
Farshid Farhat, Mohammad Kamani, Sahil Mishra, James Wang
ACM Multimedia 2017, Mountain View, CA, USA
(Acceptance rate = 64/495 = 12.93%)

ABSTRACT: Retrieving photography ideas corresponding to a given location facilitates the usage of smart cameras, where there is a high interest among amateurs and enthusiasts to take astonishing photos at anytime and in any location. Existing research captures some aesthetic techniques such as the rule of thirds, triangle, and perspective, and retrieves useful feedbacks based on one technique. However, they are restricted to a particular technique and the retrieved results have room to improve as they can be limited to the quality of the query. There is a lack of a holistic framework to capture important aspects of a given scene and give a novice photographer informative feedback to take a better shot in his/her photography adventure. This work proposes an intelligent framework of portrait composition using our deep-learned models and image retrieval methods. A highly-rated web-crawled portrait dataset is exploited for retrieval purposes. Our framework detects and extracts ingredients of a given scene representing as a correlated hierarchical model. It then matches extracted semantics with the dataset of aesthetically composed photos to investigate a ranked list of photography ideas, and gradually optimizes the human pose and other artistic aspects of the composed scene supposed to be captured. The conducted user study demonstrates that our approach is more helpful than the other constructed feedback retrieval systems.

Please cite our paper if you are using our professional portrait dataset.

@inproceedings{Farhat:2017:IPC:3126686.3126710,
author = {Farhat, Farshid and Kamani, Mohammad Mahdi and Mishra, Sahil and Wang, James Z.},
title = {Intelligent Portrait Composition Assistance: Integrating Deep-learned Models and Photography Idea Retrieval},
booktitle = {Proceedings of the on Thematic Workshops of ACM Multimedia 2017},
series = {Thematic Workshops ’17},
year = {2017},
isbn = {978-1-4503-5416-5},
location = {Mountain View, California, USA},
pages = {17–25},
numpages = {9},
url = {http://doi.acm.org/10.1145/3126686.3126710},
doi = {10.1145/3126686.3126710},
acmid = {3126710},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {deep learning, image aesthetics, image retrieval., photographic composition, portrait photography, smart camera},
}

 

Optimal Scheduling in Parallel Programming Frameworks

FORK-JOIN QUEUE MODELING AND OPTIMAL SCHEDULING IN PARALLEL PROGRAMMING FRAMEWORKS

ABSTRACT

MapReduce framework is widely used to parallelize batch jobs since it exploits a high degree of multi-tasking to process them. However, it has been observed that when the number of servers increases, the map phase can take much longer than expected. This thesis analytically shows that the stochastic behavior of the servers has a negative effect on the completion time of a MapReduce job, and continuously increasing the number of servers without accurate scheduling can degrade the overall performance. We analytically model the map phase in terms of hardware, system, and application parameters to capture the effects of stragglers on the performance. Mean sojourn time (MST), the time needed to sync the completed tasks at a reducer, is introduced as a performance metric and mathematically formulated. Following that, we stochastically investigate the optimal task scheduling which leads to an equilibrium property in a datacenter with different types of servers. Our experimental results show the performance of the different types of schedulers targeting MapReduce applications. We also show that, in the case of mixed deterministic and stochastic schedulers, there is an optimal scheduler that can always achieve the lowest MST.

 

KEYWORDS

Stochastic processes, Computational model, Delayed Tailed Distribution, Optimal scheduling, Cloud computing, Synchronization, Queuing Theory, MapReduce, Stochastic Modeling, Performance Evaluation, Fork-Join Queue.

Modeling and Optimization of Straggling Mappers

ABSTRACT
MapReduce framework is widely used to parallelize batch jobs since it exploits a high degree of multi-tasking to process them. However, it has been observed that when the number of mappers increases, the map phase can take much longer than expected. This paper analytically shows that stochastic behavior of mapper nodes has a negative effect on the completion time of a MapReduce job, and continuously increasing the number of mappers without accurate scheduling can degrade the overall performance. We analytically capture the effects of stragglers (delayed mappers) on the performance. Based on an observed delayed exponential distribution (DED) of the response time of mappers, we then model the map phase by means of hardware, system, and application parameters. Mean sojourn time (MST), the time needed to sync the completed map tasks at one reducer, is mathematically formulated. Following that, we optimize MST by finding the task inter-arrival time to each mapper node. The optimal mapping problem leads to an equilibrium property investigated for different types of inter-arrival and service time distributions in a heterogeneous datacenter (i.e., a datacenter with different types of nodes). Our experimental results show the performance and important parameters of the different types of schedulers targeting MapReduce applications. We also show that, in the case of mixed deterministic and stochastic schedulers, there is an optimal scheduler that can always achieve the lowest MST.

[Tech Report] [Master Thesis] [IEEE Trans]

Last version > MapReduce_Performance_Optimization

Image Steganalysis of Low Rate Embedding in Spatial Domain

Abstract

LSB embedding in spatial domain with very low rate can be easily performed and its detection in spite of many researches is very hard, while BOSS competition has been held to break an adaptive embedding algorithm with low rate. Thus, proposing powerful steganalyzer of very low rate in spatial domain is highly requested. In this thesis it has been tried to present some algorithms to detect secret message with very low rate in spatial domain using eigenvalues analysis and relative auto-correlation of image.

First approach is based on the analysis of the eigenvalues of the cover correlation matrix that we used for the first time. Image partitioning, correlation function computation, constellation of the correlated data, and eigenvalues examination are major challenging parts of our analysis method. The proposed method uses the LSB plane of images in spatial domain, extendable to transform domain, for detecting low embedding rates that is a major concern in the area of the LSB steganography. Simulation results show that the proposed approach improves over some well-known LSB steganalysis methods, specifically at low embedding rates.

Our second image steganalysis method suggests analysis of the relative norm of the image parts manipulated in an LSB embedding system. Image partitioning, multidimensional cross-correlation, feature extraction, and rate estimation, as the major steps of the main analysis procedure. We then extract and use new statistical features, Parts-Min-Sum and Local-Entropies-Sum, to get a closer estimate of the embedding rate and the detection performance. Our simulation results, as compared to some recent steganographic methods show that our new approach outperforms some well-known, powerful LSB steganalysis schemes, in terms of true and false detection rates.

Keywords: Image Steganalysis, Eigenvalues Analysis, LSB Embedding, Relative Autocorrelation, Parts Min Sum, Embedding Rate Estimation, Local Entropies Sum.

Image Steganalysis of Low Bit-rate Embedding

Optical CDMA Network Simulator (OCNS)

Optical CDMA Wireless Multi-User Network System includes some transmitters and receivers. In this network, an Optical Orthogonal Code (OOC) is assigned to each user (Tx or Rx) to connect to its equivalent-OOC user and after synchronization between this two equivalent-OOC user, they can send and receive data to/from each other.
In this project, I worked to design and Implement a simulator for Optical CDMA Wireless Multi-User Network. This simulator has eliminated some of practical problems like number of users can be used by network practically.
OCNS is the name of the simulator for Optical CDMA Networks. I did this project as my BS Project. My supervisor, Prof. Pakravan, suggested me this project in April 2004. In July 2004, I finished the documentation of this project in persian language. I developed OCNS by using Visual C++ software. I’ve presented the defined classes in my project below.

 

Defined Classes:
CAboutDlg
CBit
CBuffer
CChildFrm
CChip
CCode
CCounter
CCRC
CData
CDataDialog
CFIR
CFIRDialog
CFrame
CGetNumDialog
CHeader
CMainFrame
CMedium
CMediumDialog
CMSFlexGrid
COCNSApp
COCNSCntrlItem
COCNSDoc
COCNSView
CResource
COleFont
CPicture
CRowCursor
CRx
CRxDialog
CSim
CSimDialog
CSimShowDialog
CStdAfx
CTx
CTxDialog
CTxRx

 

References:

[1] Farshid Farhat, “Optical CDMA Network Simulator,” BS Thesis, Sharif University of Technology, 2005.